That's a nice approach....
But as you mentioned, tweaking of the buffers (so long as it's possible/effective) would be ideal.
Hmmm, I've no idea about that side of things, wish there was someone that does

I would be surprised if tweaking buffers on the transmitter would have much impact. What I imagine you need is extra buffers in the receiver.

What appears to be happening is that the wlan scan is consuming the antenna for a period of time. During this time, no bluetooth audio can be transmitted, so the bluetooth receiver experiences a gap. If the receiver does not have a sufficient buffer, then there is a dropout.

If the receiver could have a larger buffer, and delay the stream, then it could potentially cope with the tranmission gap. However the receivers tend to be fairly basic devices without much configurability. Unless there is a way with the a2dp protocol to tell it to have a longer delay and/or bigger buffer. I don't know much about the protocol however, and I doubt this can be done.

What ought really to happen is that the transmitter (N9) should understand that when a user initiates a bluetooth audio stream, that they are indicating that bluetooth should have priority over wlan. Then the wlan processes should act in a way not to interfere with the bluetooth stream.

Perhaps a way to do this is to have an app which automates rainisto's solution, so that the wlan functionality can be temporarily disabled when a bluetooth stream is running, and restored when it is no longer required. Would that be hard to do ?

EDIT: The Gconfik tool is that already I suppose, although it requires remembering the path to the parameter to change ...